Prompts

Productivity prompts for Claude worth stealing

Eight prompts, each one a brief rather than a wish: fixed output structure, a flag-don't-guess clause, and a length limit. They cover meetings, capture, review, triage, decisions, daily planning, email, and kickoffs, and every one has survived months of daily use across weekly business builds. Copy them as written; adjust the nouns.

A prompt is productivity infrastructure only if you'll still be using it in a month, so the filter for this list was survival, not cleverness. The pattern you'll see in all eight: name the structure you want back, cap the length, and tell Claude what to do with ambiguity (flag it) so you never get confident inventions.

Which prompts handle meetings?

1. The meeting structurer. Paste your rough fragments right after the call:

Here are my rough notes from a meeting. Structure them into three sections: Decisions (statements of what was agreed), Action items (a Markdown checklist, each with an owner and a due date if one was mentioned), and Open questions. Resolve my shorthand where obvious, flag anything ambiguous instead of guessing, and keep it brief.

The workhorse, and the full workflow around it, including the no-notes debrief variant, is in the meeting notes guide.

Which prompts capture what you'd otherwise lose?

2. The end-of-session capture. The last thing I type in any real working conversation:

Before we wrap: review this conversation and list (1) decisions we made, as plain statements, (2) tasks that emerged, as a Markdown checklist, (3) anything I said I'd think about or follow up on. Format it so I can paste it straight into a notes file. Skip pleasantries and don't add items we didn't discuss.

3. The decision log. For the moment something big gets settled mid-conversation:

We just made a decision. Write a decision-log entry for it: date, the decision in one sentence, the options we considered, why we chose this one, and what would make us revisit it. Five lines maximum, no preamble.

These two are the prompts I eventually automated, because typing them is exactly the friction that kills capture habits by Thursday. The automated version is JotThread: /jot does prompt 2's job in one command, mid-conversation, filed by project into a file you own. The free-prompt path is real, though, and the capture guide covers it fully; start free and see if the habit sticks.

Which prompts run your reviews and planning?

4. The weekly review. Paste your notes file, task list, or both:

Here's my notes-and-tasks file. Run my weekly review: (1) what got done this week, (2) what's stalled and your best guess why, based only on what's in the file, (3) what's genuinely due or time-sensitive next week, (4) three candidate priorities for the week, with one line of reasoning each. Be direct; if the file suggests I'm avoiding something, say so.

5. The daily plan. Morning, one paste of today's calendar and open tasks:

Here are today's meetings and my open tasks. Build today's plan: pick the one task that matters most and slot it into my largest free block, fit two smaller tasks around the meetings, and tell me what I should explicitly NOT attempt today. Respect the meetings as fixed. Output as a simple timeline.

The "not attempt" clause sounds gimmicky and is the most useful line in it; an honest cut list is what makes the rest of the plan real.

Which prompts sort the incoming?

6. Task triage. For the overwhelmed-list moment:

Here's my task list. Triage it: (1) quick wins under 15 minutes, (2) the one or two tasks that actually move my main goal, which I'll state below, (3) tasks to delegate or drop, with a one-line reason each, (4) anything that's really a project pretending to be a task; break the first one into steps. My main goal right now: [goal].

7. Email-to-actions. Paste a long thread instead of rereading it three times:

Here's an email thread. Extract: (1) what's being asked of me, exactly, (2) commitments others made, with names, (3) a Markdown checklist of my resulting actions, (4) a draft reply under 120 words that confirms next steps. Flag anything where the thread is genuinely unclear rather than papering over it.

Which prompt starts projects properly?

8. The kickoff brief. Ten minutes at the start of anything new:

I'm starting a new project: [one-paragraph description]. Interview me, one question at a time, until you can write a one-page brief covering: goal, definition of done, constraints, first three concrete tasks, and the main risk. Push back where my answers are vague. Then output the brief in Markdown so I can save it as the project's founding file.

The interview format matters: answering questions surfaces what you actually don't know yet, which is the entire point of a kickoff. The brief then becomes the project's first file, and if you work in Claude Projects, its founding knowledge; the Projects guide covers that setup.

Common questions

Should I save these as custom instructions?

No; custom instructions apply everywhere, and these are job-specific. Keep them one paste away, or promote the ones you use daily into a skill so they trigger themselves instead of waiting to be typed.

Do these work in other AI tools?

Mostly yes; they're structured briefs, not model tricks. The capture prompts assume the conversation holds the material, which is true in any chat AI you genuinely work in.

Why do they all say "flag instead of guessing"?

Because the failure mode of AI productivity isn't wrong answers, it's confident inventions: a due date you never set, a decision you never made. One clause converts those into flagged questions, which is the difference between a tool you check and a tool you trust.

Prompts 2 and 3, as one command.

/jot captures notes, decisions, and tasks mid-conversation, filed by project in a Markdown file you own. $29 once, no subscription.

Get JotThread · $29